Computer scientists from University of Nottingham and Kingston University have made a revolutionary breakthrough in the field of vision and graphics research. They have found a way to produce 3D facial reconstruction using only one 2D image.
The developers' app is able to transform full-coloured images into 3D models of users' faces within seconds. It has been hugely popular so far, with over 400,000 people giving it a try.
The research is called 'Large Pose 3D Face Reconstruction from a Single Image via Direct Volumetric CNN Regression'. It is conducted by Aaron Jackson and Adrian Bulat, both from the Computer Vision Laboratory in the School of Computer Science. The project is overseen by Georgios Tzimiropoulos, Assistant Professor working in the School of Computer Science.
The outcome of the project will be presented at the International Conference on Computer Vision (ICCV) 2017 taking place in Venice in the coming month.
Whilst the team has yet to perfect their technology, this is already a much anticipated breakthrough among computer scientists. The technique relies on the Convolutional Neural Network (CNN) - a branch of AI which allows computers to learn on their own rather than being specifically programmed.
"The main novelty is in the simplicity of our approach which bypasses the complex pipelines typically used by other techniques. We instead came up with the idea of training a big neural network on 80,000 faces to directly learn to output the 3D facial geometry from a single 2D image," said Dr Tzimiropoulos.
This is a hugely complex problem. Existing systems would require several photos and would encounter various difficulties. These are connected to factors such as facial expressions, poses, and lighting.
Aaron Jackson explained, "Our CNN uses just a single 2D facial image, and works for arbitrary facial poses (e.g. front or profile images) and facial expressions (e.g. smiling)."
Adrian Bulat added, "The method can be used to reconstruct the whole 3D facial geometry including the non-visible parts of the face."
Their research employs technologies related to deep learning, in which machines can utilize artificial neural networks to piece together information just as the human brain does.
Dr Vasileios Argyriou, from Kingston University's Faculty of Science, Engineering and Computing, shared: "What's really impressive about this technique is how it has made the process of creating a 3D facial model so simple."